Thanks for this PR! I noticed that in the cupy installation instructions that you shared, it is recommended to use the following command conda install -c conda-forge cupy. It is mentioned that cudatoolkit will be installed automatically this way.

Is there a reason that you prefer using the default conda channel for the main installation instructions?

In addition, I would add the following command in the installation instructions: conda install -c conda-forge cupy cuda-version=11.8 to provide an example on how to manually select the CUDA version. Otherwise, the comment below the installation command becomes obsolete as well.

Adding screenshots of each step can be very useful for new users as most of them are not used to working on Unix based systems